Glimpse at the Dawn of AI Industrialization Development: How HUAWEI CLOUD Carries AI Through to the End

Glimpse at the Dawn of AI Industrialization Development: How HUAWEI CLOUD Carries AI Through to the End

When we discuss the challenges of integrating AI into the industry, in fact, to a large extent, we will face the mismatch between AI technology and the enterprise’s own production system, production process and automation system.

In industry discussions, intelligence is often closely associated with automation. However, in the actual process of enterprise deploying AI, it is found that AI technology is often independent of the enterprise automation system, and even a large number of manual parameter adjustment, training, model adaptation and other tasks will reduce the degree of automation of the enterprise.

Logically speaking, AI should obviously not become a burden on enterprises, nor should it become a technical proposition contrary to enterprise automation. Whether “intelligence” and “automation” can be combined to create a service that liberates productivity with enterprise users as the main body has become the key to the transformation of AI technology from “towards industrialization” to “in-depth industrialization”.

On June 18, Huawei Cloud TechWave Global Technology Summit (Artificial Intelligence & Data) was held in Shanghai. During the period, HUAWEI CLOUD demonstrated the strategy of continuously investing heavily in the fields of data and artificial intelligence, and released a series of new AI products, and announced the large-scale commercial use of a series of new data products.

Let’s start with the new AI series product released at the meeting – Huawei Cloud Intelligent Process Robot Solution, and talk about Huawei Cloud’s strategy and determination to “carry out AI to the end”.

Digital employees, the only way for enterprise automation

The reason why RPA is important is that it can provide enterprises with “digital employees” who can actually work although they are invisible and intangible, and specialize in handling the ubiquitous process work in the enterprise.

In all kinds of enterprises, large and small, it is believed that there are highly repetitive, process-oriented, and basically uncreative tasks, such as invoice entry, meter reading, and document filling. However, these tasks support the daily operation of the enterprise and cannot be given up categorically.

So can these process tasks be handed over to machines? In the information age, this problem was discovered very early. As early as the 1990s, the prototype and base technology of RPA had already appeared. In 2016, Gartner released a report pointing out that the full-blown RPA has been able to replace manual processing of complex, cumbersome, and heavy affairs, thereby reducing the labor cost of enterprises and improving overall work efficiency.

Although the vision of helping companies create “digital employees” is good, there are also a series of limitations and drawbacks. For example: RPA is difficult to handle flexible process work; most RPA services can only handle standardized data, lack of non-standardized data processing capabilities, complex work is difficult to handle; RPA also lacks knowledge learning ability and environmental adaptability, and real work needs Mismatch.

The existence of these problems has limited the development of RPA. Facing these challenges, HUAWEI CLOUD aims at the combination of intelligence and automation. At the summit, HUAWEI CLOUD released the intelligent process robot solution, combining RPA and AI technology, so as to turn “digital employees” into “smart digital employees”.

Glimpse at the Dawn of AI Industrialization Development: How HUAWEI CLOUD Carries AI Through to the End

(Zhu Haipei, Director of HUAWEI CLOUD Architecture and Technology Planning Department, introduced the HUAWEI CLOUD intelligent process robot solution)

Different from the past, the new HUAWEI CLOUD intelligent process robot solution mainly has the following features:

1. Sensitive and eager to learn: integrate the overall skills of AI into RPA.

On the whole, HUAWEI CLOUD has created an AI+RPA system that has the ability to learn and can disassemble unstructured data in traditional business, allowing RPA to adapt to financial, tax, IT, business, procurement, HR, project Management, salary management and other more complex and specific business processes.

2. Smart eyes and ears: Integrate high-quality AI capabilities with RPA.

The HUAWEI CLOUD intelligent process robot solution incorporates HUAWEI CLOUD’s leading AI capabilities such as OCR and NLP, and can achieve high precision and low error in key capabilities such as recognition and understanding. For example, in a joint venture with a bank, the HUAWEI CLOUD intelligent process robot solution has improved the efficiency by 450 times and reduced the error rate by 90% for bank bills of lading and letter of credit identification and automatic entry.

3. Know what you hear and know ten: Low-code capabilities are integrated into RPA.

In addition to AI, HUAWEI CLOUD also integrates low-code development capabilities into RPA, so as to achieve a more accurate and efficient match between RPA and business processes and work links. It not only ensures the lower entry barriers for enterprises, but also ensures the feasibility of customized development.

As an important strategic partner in Huawei’s cloud ecosystem and a pioneer in the field of RPA, iSoftStone has already taken the lead in applying the AI+RPA technology jointly developed by the two parties to internally improve process efficiency and optimize process quality, and conduct its own process automation practice. At present, iSoftStone has used the HUAWEI CLOUD intelligent process robot solution in 273 business scenarios. In 2020, it will process 2 million+ documents, increase the processing efficiency by 6.5 times, and save nearly 8.8 million yuan in annual costs.

It is not difficult to see that enterprises with complex processes, more unstructured data, and greater business flexibility can obtain a more stable, reliable, intelligent, and automated experience from the HUAWEI CLOUD intelligent process robot solution.

On the same day, HUAWEI CLOUD also released the city intelligence center. By building an urban AI development and operation platform, it provides three major engines: global perception engine, knowledge computing engine, and robot engine, so as to realize cross-modal, cross-departmental collaboration and global precision of urban conditions. Perception helps realize intelligent urban governance and precise services.

Glimpse at the Dawn of AI Industrialization Development: How HUAWEI CLOUD Carries AI Through to the End

Zhao Mingming, vice president of Digital Zhengtong, said that Digital Zhengtong and Huawei Cloud have jointly created multiple solutions such as the 12345 government affairs hotline intelligent perception platform, city operation one-network unified management, urban comprehensive management service platform, urban social governance and city-level parking, etc. Provide a high level of service.

“One use, three uses”, carry AI to the end

In fact, looking at Huawei Cloud’s overall strategy in the AI ​​field in recent years, we will find that Huawei Cloud hopes to continue to be application-oriented, and gradually deepen AI technology into the core of the enterprise, so that AI can unleash productivity.

As Shi Jilin, vice president of Huawei Cloud BU and president of global marketing and sales services, said during the summit: “AI needs to start from a single point of innovation, and truly enter the main business process and main production process of the enterprise, so that it can carry out system innovation. Enter After reaching this stage, we will find that our thinking about AI is not only to improve the efficiency or competitiveness of a certain scenario of the enterprise, but to reconstruct the main business process of the enterprise and the core competitiveness of the enterprise.”

Glimpse at the Dawn of AI Industrialization Development: How HUAWEI CLOUD Carries AI Through to the End

(Shi Jilin, Vice President of Huawei Cloud BU, President of Global Marketing and Sales Services)

In order to help AI enter the core system of the enterprise, increase the penetration rate of AI technology, and control Huawei Cloud’s overall layout and industrial actions in the AI ​​field, I summarize it as a strategic structure of “one use, three transformations”. “One use” refers to enterprise application as the guide for all innovations, and directs all AI values ​​to be implemented; while “three modernizations” refers to simplification, industrialization, and sceneization. Under these three paths, HUAWEI CLOUD continues to improve the AI ​​product architecture and integrate AI into the core system of the enterprise.

1. Simplify.

Development is the most basic requirement for the implementation of AI, and HUAWEI CLOUD’s first starting point in advancing the AI ​​track is to make AI development easier for AI development platform ModelArts. It transforms industry knowledge into assets through one-stop AI development capabilities and rich industry suites. ModelArts3.0 has new features such as fusion of backbone models, federated learning, model intelligent evaluation and diagnosis, and high-performance AI computing, which can accelerate the entry of AI into the production system. It can be said that the simplification of AI development will continue to advance step by step.

Simplified AI development and deployment has accelerated the intelligentization process of many industries. For example, Malanshan Video Cultural and Creative Industrial Park has developed AI automatic coloring models and services, AI iterative optimization process tools based on HUAWEI CLOUD AI-enabled ModelArts and cloud clusters. Large-scale AI video coloring production line.

2. Industrialization.

The core of AI industrialization development is to improve the standardization level in the early stage and lower the customization threshold for enterprises. Among them, the pre-trained large model is the core solution in the industry. In April this year, HUAWEI CLOUD released the Pangu large model, which has accumulated a large amount of industry knowledge and data, and can realize the generalization, generalization, and large-scale replication of an AI large model in many scenarios. At present, the industry’s first 100-billion-level generation and understanding Chinese NLP large model has been released, and the industry’s largest CV large model contains more than 3 billion parameters, which truly drives AI to the era of industrialization.

3. Sceneization.

AI solutions specific to industry scenarios are the main channel for AI to enter the industry. At present, HUAWEI CLOUD has launched many solutions with partners in different scenarios in the fields of industry, Internet, and e-commerce to realize the application of AI scenarios. Across the country, HUAWEI CLOUD has implemented 600+ AI projects in more than 10 industries.

In the coal industry, HUAWEI CLOUD EI industrial intelligence combines the coal blending mechanism model with AI to help Shiheng Special Steel complete the “AI optimized coal blending”. At present, this capability has supported stable commercial use, and its coke quality prediction accuracy rate exceeds 97%, helping enterprises reduce the coal cost per ton of coke production by an average of about 15 yuan.

In the field of medicine, the Chinese Academy of Sciences Shanghai Pharmaceuticals and the Huawei Cloud team have launched a number of AI drug research and development cooperation projects. Among them, the large-scale drug virtual screening cloud service can achieve a ten-fold increase in drug screening efficiency by utilizing the super computing power of Huawei Cloud’s 10,000 cores.

In the automotive industry, China FAW applied Huawei Cloud knowledge computing solution to build the FAW knowledge computing platform. The one-time repair rate of the pilot 4S shop of FAW Hongqi increased by 4%, the waiting time for customer maintenance decreased by 23%, and the intervention rate of manufacturer support A 30% reduction and a 30% reduction in training time for maintenance technicians.

From these industry cases and developer practices, it can be seen that Huawei Cloud has different characteristics from most cloud computing vendors in the AI ​​field: instead of providing tools and capabilities on the technical side, it is on the industry side, and everything is based on user needs. Guided by what and what problems users need to solve, gradually build AI supply-side capabilities.

Based on the continuous advancement of simplified development, continuously lower the AI ​​threshold; take industrialization as the pre-integration as the starting point, improve the standardization of the AI ​​industry, and control the comprehensive cost; take the scenario as the guide, and go deep into the industry space. These three roads are all progressing steadily and supporting each other, so that products like ModelArts, Knowledge Computing, and Pangu Large Model break the conventional AI products, but highly meet the needs of enterprises and industries.

AI is entering the dawn of industrialization, but the sun will not bloom for no reason, but one step at a time is needed to promote AI to the industry and attract the industry to embrace AI.

The reason why HUAWEI CLOUD’s AI differentiation has been recognized by the industry is that compared to AI floating on the water, HUAWEI CLOUD is better at carrying AI to the “bottom”: the bottom layer of technology, the foundation of the industry, and the confidence of the times.

The Links:   CR6L-250/UL LM64C509